iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Keenable is a web-search API built for AI agents rather than for people scanning a results page. Its two headline claims, an index of more than 100 billion documents and a p95 latency below 250 ms in US East, are company-published figures. The material available for this article does not include an independent audit or the test methodology behind either number. This article explains what Keenable offers, how its architecture argument works, what the large-index claim does and does not imply, and what to verify before relying on it.
What Keenable is
Keenable presents itself as independent web-search infrastructure for AI labs and agents. Its core product is a Search API that returns ranked web pages along with extracted page text. Its developer materials also describe a fetch operation that returns clean markdown. In practice that suggests a two-step pattern: search to find candidate sources, then fetch to read the one an agent chooses. Keenable’s official materials describe the software and its service tiers; they do not describe a physical product.
The three product surfaces
Search API
The Search API is the baseline offering. It returns ranked results with page text, so an agent receives material it can read or pass to a model rather than only titles and snippets. Developer access is through the API directly and through the SDKs described later in this article.
Free tools Windows power users keep installed
One-click scans. No signup required.
SELECT
SELECT is a SQL-like interface for searching web results and extracting structured fields from them. A query can filter, group and aggregate the extracted fields and return a table or report. Keenable’s stated motivation is that some answers are properties of a set of pages rather than facts that appear on a single page. A question such as “how many of these people moved to a new employer” is an example of that kind of answer.
#1 Best Overall
Keenable’s own SELECT essay argues that conventional ranked links and short snippets suit a person who opens one result. An agent may instead need the distribution across many pages plus structured fields. The essay offers one dated example: a report listing 46 researcher moves across 11 frontier foundation-model labs, covering January 2025 to August 2026. That example shows the output format. It is not a general quality statistic, and the essay is the company’s own design argument rather than an independent comparison.
Time Machine
Time Machine performs point-in-time search over prior versions of pages. The query time sets both the historical corpus being searched and the ranking applied to it. Keenable’s official site labels this capability as early access. Confirm availability on the live site before planning a workflow around it, because early-access features can be limited to certain accounts or regions.
Pages or documents: reading the 100B figure
The title uses “100B-Page,” but Keenable’s official materials generally describe the indexed units as documents. The company’s homepage states an index of 100B+ documents. The material does not define what counts as a document, for example whether a versioned page, a file or a page revision is counted separately. Treat “100B pages” as informal shorthand for the company’s own wording, and read the count as the company’s unit rather than a verified page total.
The architecture argument behind the index size
Keenable’s central argument is about cost. Serving and scanning the whole web for every query is expensive, so a search system has to narrow the candidate set quickly and according to the query. CEO Andrey Styskin, Keenable’s co-founder and CEO, put it this way in a TechCrunch report dated August 25, 2026:
“If you do not fine-tune your index structures for a specific task, the cost of serving and scanning the whole internet is enormous because of the volume. That’s why you need to innovate on how you can narrow the search space based on your query very fast. This is what we are bringing to the table.”
The trade-off is straightforward. Broad coverage helps an agent find sources it would otherwise miss. Broad coverage also raises the cost of each query unless the system can cut the candidate set early. Keenable’s answer is index structure plus query-driven narrowing. That is a design claim, and the company has not published a breakdown showing how much narrowing contributes to its latency or price.
For an evaluation, that means index size alone is a weak quality measure. A more useful set of questions is:
- Does the index contain the sources your workload needs, and how current are they?
- Does the search return relevant results for your queries rather than for a generic sample?
- How complete and clean is the extracted text?
- What do the latency distribution and the price per useful answer look like at your volume?
Latency: what the p95 figure does and does not say
Keenable’s official homepage states p95 latency “less than 250 ms” in US East. The claim names a geography and a percentile, which is more than many vendors publish. It does not disclose the test setup, the query mix, the client’s location relative to US East, whether network time is included, or whether fetch calls are counted.
If latency matters for your agent, measure it under your own conditions:
Rank #4
- Run the client from the same region where your agent will run.
- Use a sample of queries that resembles your real workload, not a single repeated query.
- Record p50 and p95 separately, since the median and the tail answer different questions.
- Include end-to-end time from request to parsed response, and time the fetch step separately if you use it.
Pricing
Keenable publishes tiered pricing on its pricing page. The figures below were observed in search results, and the pricing page was last crawled several weeks before this article was prepared. Check the live page before making a purchase decision.
| Tier or offer | Published price | Access and deployment | Qualifiers to verify |
|---|---|---|---|
| Agent Builder | $4 per 1,000 requests | Cloud only, pay as you go | Current terms on the live pricing page |
| Frontier | $1 per 1,000 requests at 100 RPS or more | Dedicated capacity for AI labs and inference platforms; cloud and on-premises | Rate threshold, eligibility and terms; not confirmed as current in the observed listing |
| Free offer | 100,000 requests a month | Not stated | Eligibility and terms; not confirmed as current in the observed listing |
At the observed Agent Builder rate, a workload of 50,000 requests a month would cost about $200 at list price before any free allowance. The listing does not confirm whether the free allowance applies to that tier, so do not assume it does.
Integration paths
- Python and TypeScript SDKs: The SDK documentation describes keyless defaults. An optional API key affects rate limits, so a prototype can start without a key and a production system should plan for one.
- LangChain integration: Keenable lists a LangChain integration for agents built on that framework.
- MCP server: The MCP repository documents hosted search and fetch tools and a keyless request cap. Check the repository for the current cap value, since package versions and defaults can change.
These integrations show how an agent can call search and fetch. They do not establish reliability or how the developer experience compares with other providers.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Adoption and partnerships
The same TechCrunch report says Keenable stated that its API is in production at several AI labs and inference providers, used for both training and runtime. The customers were not named. The report also names a partnership with Gradium, a voice AI company, for live information retrieval. These are Keenable’s statements as reported. A deployment at unnamed customers shows that the service is in use. It does not show that Keenable outperforms competing services on any particular workload.
How to evaluate Keenable against alternatives
Keenable’s materials do not include a head-to-head comparison with named competitors on every axis. The following are sensible evaluation axes, inferred from the product surfaces and the trade-off the company describes:
- Corpus coverage and freshness for your target sources
- Result relevance on your own query set
- Extraction completeness, including whether structured fields come back accurately
- p50 and p95 latency, with region and test setup recorded
- Cost per request at your expected volume and rate limits
- Historical snapshot support, if you need point-in-time answers
- SDK, LangChain and MCP integration fit with your stack
- Deployment and data-control options, including whether on-premises is available on your tier
- Whether benchmark results are independently reproducible
What is and is not established
- Established: Keenable offers a Search API with extracted page text, a fetch operation, SELECT, and Time Machine (early access), plus Python, TypeScript, LangChain and MCP integrations.
- Company-published, not independently verified: the 100B+ document index, the p95 latency under 250 ms in US East, and the benchmark chart on the homepage, which Keenable presents as its own quality measure.
- Reported, not independently confirmed: production use at several AI labs and inference providers, and the Gradium partnership.
- Not established: comparative performance against other search APIs, the test methodology behind the latency and benchmark claims, and the exact definition of a document in the index count.
The Bottom Line
Keenable is a credible option to test if your agents need broad web search with extracted text, and its SELECT and Time Machine surfaces address needs that plain ranked links do not. Treat the index size, latency and benchmark figures as the company’s claims until your own measurements confirm them for your queries, region and volume.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
